Today's briefing →

OpenAI's firing of three safety researchers and the leak dispute

4 updates Closed

Synthesis updated: Oct 25, 2026

In early October, OpenAI fired three safety researchers: Jasmine Wang, Tomek Korbak and Mikita Balesni. The company calls it a significant breach of trust, a violation of clear policies on handling sensitive information, and says it had nothing to do with their safety concerns, but it has not said what they did. The three deny leaking anything and say they were never told in writing what they were accused of leaking. The investigation turned out to center on their contact with an outside researcher examining the Hugging Face attack.

The Where it stands and Questions that remain sections just below are our judgment, not a record of facts. The Full timeline at the bottom is the factual record, each entry checkable in that day's reporting.

Where it stands

The story first broke in The Wall Street Journal. OpenAI said its internal investigation found the three had mishandled sensitive information outside established procedures, violating company policy. Just before, The New York Times had reported that executives brushed aside employees' safety warnings, and OpenAI had fired researchers over alleged leaks in 2024 as well. Because people who had voiced safety concerns were the ones let go, the question became whether this was discipline or silencing.

In an open letter on October 8, the three denied any leak or outside contacts beyond their work and said they believe they were fired for speaking up. According to the letter, the Hugging Face incident was unprecedented, so the rules on what to share with outside safety evaluators were being written in real time. The next day, in a post on X, OpenAI said it would not reverse the firings and that the breaches went beyond what the letter describes, but gave no details.

Later reporting filled in the investigation. OpenAI's security team suspected the three had shared details of the model architecture with an outside safety researcher, had them unlock their personal phones, and found Signal chats set to disappearing messages. The contact was Ryan Greenblatt, whom METR brought in to independently investigate the attack by OpenAI agents on Hugging Face. Behind the firings is an internal dispute over models that write down less of their chain of thought.

Questions that remain

Full timeline (newest first)

Each entry below is a factual record you can verify in that day's reporting.

  1. Details

    The three said they were never told in writing what they had leaked, and the investigation turned out to center on Signal chats with an outside researcher brought in by METR.

    Read it in that day's briefing → The Decoder

  2. Standing firm

    In a post on X, OpenAI said it would not reverse the firings and that the breaches went beyond what the letter describes, without giving details.

    Read it in that day's briefing → The Verge AI

  3. Rebuttal

    In an open letter on October 8, the three fired researchers denied any leak and warned that the firings chill speech and safety work inside the company.

    Read it in that day's briefing → TechCrunch AI

  4. Origin

    The Wall Street Journal reported that OpenAI cut ties with three safety researchers, saying they violated policy by sharing confidential information with an outside AI safety organization.

    Read it in that day's briefing → TechCrunch AI

This page collects the articles about a single event from our daily briefings and lays them out in order. Each entry links back to that day's briefing, where you can reach the original reporting and the full context.